From: John Tserkezis 
Reply-To: Fidonet AVtech Echo 

Bob Lawrence wrote:

>>Frequency meters measure *both* period and frequency.

> JT> There are hardware implications if doing both, (easy to do, but
> JT> still adds a little to hardare costs). If you're using a
> JT> microprocessor, there are no differences in hardware, but major
> JT> differences in software. 

>  There is a hardware variable. You need to time a CYCLE which needs
> an accurate level-sensing switch, or count cycles in a fixed TIME,
> which means the switch is less stringent. An ordinary Schmitt Trigger
> is good for 1000:1 or 0.01%. If you count 1000 cycles you get 1 ppm.
> That's why they prefer to count cycles rather than microseconds (or
> whatever).

  Depends on how the pulse was received I suppose, if it's iffy, you're 
likely to get a fair amount of jitter, which would fluctuate the 
readings by a good amount (it might average OK, but the jitter would 
make it useless for human readability).
  I suppose you could average the time interval between several pulses, 
it would slow the update somewhat, but would still be much faster than 
just plain pulse counting.

>>It's insane to measure frequency rather than time at that speed,
>>especially when you only need three digits. I despair of
>>enginerring in Australia today...

> JT> It's cheaper. (in engineering costs, not hardware- hardware
> JT> costs the same either way).

>  It's not. The problem is accuracy, and three digits is shit-easy.
> That's only 0.1% and the cheapest-crap Schmitt can do 0.01% (80dB
> S/N). What is is...is plain stupidity, and there's a lot of it about.

  Again, it is, these issues can be fixed in software.

> JT> If you were to measure period, you would need to number crunch,
> JT> as well as have a separate timer to handle the update rate. (so
> JT> the update rate is a constant, separated from the current
> JT> speed).

>  The update rate is a human-thing. Constantly changing numbers
> confuse us because our brain actually runs a little behind real-time. 

> How does it cost any extra?  In a micro, there is the clock and a
> timer. It's just a matter of finding a convenient pulse at 100ms or
> so... and update the count on that.

  Many micros have built-in timers that would suit this purpose.  Even 
if it doesn't you can still do it in software.

>  It's just a matter of whether you
> count known short-pulses in the interval between speed-pulses, or
> count speed-pulses in a long known interval. You just switch them
> around in the counter!

  Yeah, but a non-micro solution won't do the number crunching for you 
if you're measuring the time interval.

>  The number-crunch is just an inversion which can be done at 12MHz or
> something... really fast!

  Not even that.  You don't have to look at EVERY pulse that comes in, 
you can afford to skip a few and not lose anything as far as the update 
rate goes anyway.

>  A graduate is not an engineer. An engineer is someone who can solve
> a problem cheaply and reliably.

  Bullshit.  An engineer is anyone who calls themselves an engineer. 
Some even have fancy pieces of paper with the word "engineer" on it.
  Still doesn't ensure they have any specific capability.

>  I was talking to the guy in the fish shop last week, he's building
> a house, and he pays carpenters $40/hr! That's mad! You can get an
> engineer for that!

  Who cares?  Plumbers still make more than that.


--- ifmail v.2.15